2.2606.115

New Features

• OpenFiles built-in AI service (ready right after sign-in): signing in to your OpenFiles account automatically grants an "OpenFiles built-in" provider — without configuring any vendor keys yourself, use all AI capabilities: chat, AI image generation, AI video generation, AI voiceover (TTS), and voice cloning are all built-in and login-ready; built-in OFC balance badge + one-click top-up (Stripe / Alipay / WeChat), with a prompt that guides you to top up when credits run low; the model selector shows two tracks, "OpenFiles built-in" and "My Providers".
• AI video generation (enhancements): supports the full text-to-video pipeline; supports multi-keyframe sequences (N keyframes → N−1 interpolated segments auto-stitched together) and video character consistency (keyframes lock the subject); outputs embed an AIGC compliance marker.
• AI image generation enhancements: reuse the last random seed for reproducibility; controlled multi-view of the same subject (e.g. six-view e-commerce shots); semantic reference images driving consistency (character / product / scene references); generated images embed AIGC metadata + an "AI-generated" badge on the card.
• AI voiceover / speech synthesis: text-to-speech lets you pick the voice, language, and speed with instant preview; multi-speaker dialogue scripts keep each speaker's voice / accent consistent throughout (segment-by-segment synthesis + auto-stitching); audio outputs embed AIGC compliance metadata + a badge.
• Voice cloning (voice replication): upload a voice sample to clone a custom voice, gated by a built-in authorization confirmation (per the Measures for Labeling AI-Generated Synthetic Content); cloned voices can be used directly in voiceovers.
• Consistency asset library: unify "character / product reference images" and "voiceover voices" into reusable assets with stable IDs, selectable with one click across image / video / voiceover generation — enabling cross-modal consistent reuse where "the same character keeps the same voice".
• Markdown editor: the WYSIWYG editing area adds find & replace; the rendered view adds zoom controls; pasted Markdown source is automatically parsed into rich-text nodes.

Fixes

• PPTX: filled in rendering details for master / layout background shapes, theme colors and custom / preset geometry, group transforms, directional arrow presets, gradient stop ordering, and text color schemes; "Fit to window" now works on any page.
• Mindmap: tree layout now arranges by path rather than per-level column spacing; fold icon stroke is constant and decoupled from connector depth.
• Multi-tab: global keyboard shortcuts and pagination shortcuts now act only on the currently active tab, avoiding accidental cross-tab triggers.

Improvements

• All of the AI generation and asset library UI copy above is synced across 5 languages (Simplified Chinese / English / Japanese / Korean / Traditional Chinese).

changelog